Improved algorithms for neo-epitope prediction in cancer
Anne-Mette Bjerregaard1*, Sofie Ramskov2, Sunil K. Saini2, Lina Such2, Amalie K. Bentzen2, Zoltan Szallasi1,4, Morten Nielsen1,3, Sine R. Hadrup2, Aron C. Eklund1*
1Department of Bio and Health Informatics or 2National Veterinary Institute, Technical University of Denmark, Lyngby, Denmark, 3 Universidad Nacional de San Martín, Buenos Aires, Argentina, 4Boston Children’s Hospital and Harvard Medical School, Boston, USA. *Corresponding authors
Personalization of immunotherapies such as cancer vaccines and adoptive T-cell therapy will depend on identification of patient-specific neo-epitopes that can be specifically targeted. Efficient methods for identification of neo-epitopes will yield insight into tumor-immune interactions and may improve personalized immune therapy. Current methods for predicting immunogenic mutated peptides, based on features such as HLA binding and gene expression, result in a low rate of validation in screens for reactive T cells. Thus, there is a need for improved methods and new data for optimizing these methods. We have previously established a platform for neo-epitope extraction and prediction based on tumor sequencing data, MuPeXI (https://www.cbs.dtu.dk/services/MuPeXI/). For several patient cohorts we are screening for broad libraries of mutation derived peptides (150-1000 per patient) extracted based on MuPeXI, and analyzing the T cell reactivity based on high-throughput screening using DNA barcode-labeled MHC multimers. This data will likely provide novel insight into determinants of immunogenicity of neo-epitopes and define new rules to optimize algorithms for neo-epitope prediction.